Our earlier work showed that the laboratory diagnosis of Pneumocystis pneumonia could be done reliably with bronchoalveolar lavage (BAL) specimens, and often with induced sputa, particularly in conjunction with a fluorescent monoclonal antibody stain. In addition, we reported that Pneumocystis polymerase chain reaction (PCR) can increase the sensitivity of detection when testing induced sputa. More recent work from Denmark suggests that PCR can successfully detect Pneumocystis even from simple oral washes. The use of oral washes would greatly benefit patients since sputum induction is often difficult and unpleasant. We have started a prospective study to look at the comparative sensitivity of induced sputa and oral washes to detect Pneumocystis. For this study, every patient scheduled to undergo sputum induction will simultaneously submit an oral wash specimen by gargling first with a salt solution. Both the oral wash specimen and the induced sputum will be stained for Pneumocystis using our standard fluorescent monoclonal antibody procedure. All specimens will then be tested, in a blinded fashion, by PCR assays for Pneumocystis. We have designed two separate assays: one utilizing primers for the major surface glycoprotein (MSG), developed here by J. Kovacs, and another using primers described by Wakefield et al. (Lancet'336: 451-453, 1990) based on mitochondrial rRNA genes. Patients who have induced sputum stains negative for Pneumocystis will proceed to have a BAL performed, when clinically indicated. Results of the BAL specimen stains have generally been regarded as the "gold standard" and thus will be used to determine which of these patients would traditionally be regarded as negative for Pneumocystis. Preliminary studies comparing the two assays using a dilution series of known positive control material indicated that both PCR assays were similar in sensitivity. Similarly, evaluation of the patient specimens thus far suggest that both assays have similar sensitivity (4/7 MSG, 3/7 Mito). The occurrence of false-negative and possible false-positive reactions is currently being investigated by looking at the clinical information available on each of these patients. Further studies are needed to more definitively establish the clinical usefulness of this technique.